Virtual Sensor Reconstruction for Fault-Tolerant Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In systems with multiple sensors, reconstructing missing or erroneous sensor data is complex and resource-intensive, making it difficult to ensure reliable control and detection of critical operating states.
Innovation Solution
A monitoring apparatus and method that identifies sensor candidates through structural and functional analysis, calculating similarity values to generate virtual sensor data that replace faulty sensor data, using a two-stage approach to simplify and expedite the reconstruction process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If statistical relationships are calculated for each sensor pairing to reconstruct failed sensor data, then the reliability of sensor data reconstruction is improved, but the computing power requirement and hardware complexity increase significantly
Solution Approach 1:
The patent segments the sensor network into hierarchical levels (individual sensors, sensor groups, and central evaluation device). Instead of calculating all possible pairwise relationships across the entire sensor network, the system divides sensors into groups and only calculates statistical relationships within each group. This segmentation reduces the total number of calculations from O(n²) for n sensors to O(k·m²) where k is the number of groups and m is the average group size, significantly reducing hardware complexity while maintaining reconstruction reliability.
Solution Approach 2:
The patent performs preliminary action by pre-calculating and storing statistical relationships (correlation coefficients, regression equations) between sensors during normal operation. When a sensor fails, these pre-computed relationships are immediately applied to reconstruct the failed sensor's data. This eliminates the need for real-time calculation of statistical relationships during fault conditions, reducing the computing power requirement at the moment of failure while maintaining high reconstruction reliability.
2Reliability
If a knowledge-based approach is used to identify sensor relationships, then the reliability of sensor data reconstruction is improved, but the complexity of creating and maintaining the knowledge base increases
Solution Approach 1:
The patent implements self-service by enabling the sensor network to automatically build and maintain its own knowledge base through statistical analysis of sensor data. Instead of requiring manual domain expertise to create relationship models, the system autonomously computes correlation coefficients and regression relationships from operational data. This eliminates the need for complex manual knowledge engineering while maintaining reliable reconstruction, as the system learns relationships directly from actual sensor behavior.
Solution Approach 2:
The patent transforms the knowledge base from a static, manually-crafted structure into a dynamic, data-driven structure. Statistical parameters (correlation coefficients, regression slopes) are continuously updated based on incoming sensor data, allowing the knowledge base to adapt to changing system conditions. This parameter-based approach simplifies the knowledge base structure to standardized statistical models that are easier to maintain and update compared to complex rule-based knowledge structures.
3Manufacturing precision
If virtual sensor data are calculated using sensors with highest similarity values, then the manufacturing precision of reconstructed data is improved, but the time required to identify suitable sensor candidates increases
Solution Approach 1:
The patent applies preliminary action by pre-computing and storing similarity values (correlation coefficients) between all sensor pairs during normal operation. When a sensor fails, the system immediately queries these pre-stored similarity values to identify the most suitable candidate sensors for reconstruction, rather than calculating similarities in real-time. This reduces the candidate identification time from O(n²) calculations to O(n) queries, while maintaining high precision by selecting sensors with the highest pre-determined similarity.
Solution Approach 2:
The patent applies local quality by focusing computational resources on identifying and using only the top-k most similar sensors for reconstruction, rather than considering all possible sensor combinations. The system ranks candidate sensors based on similarity values and selects only those with the highest scores, concentrating the reconstruction effort on the most relevant local relationships. This reduces the time complexity from evaluating all n sensors to evaluating only the top-k candidates, significantly reducing identification time while maintaining high reconstruction precision.
Data Source
AI summary
The monitoring of a technical system using sensor data. In the event of the failure of a sensor, in this case, for the failed sensor, virtual sensor data are created on the basis of the remaining functional sensors. In this case, the sensors for the calculation of the virtual sensor data are selected in two stages. In a first step, firstly, possible candidates of sensors are ascertained on the basis of a knowledge-based approach and the topology of the system. A second step involves calculating a mathematical relationship between the sensor data of a faulty sensor and the possible candidates of sensors for the calculation of the virtual sensor data. Those sensors which form a suitable basis for the calculation of the virtual sensor data can be identified in this way.